Socket A temperature guides...PLS READ ALL

[Confused]

OK, we all know that Mike has lots of knowledge on the 'inaccuracies' of, as he says, ALL socket thermistors, and how they do not show the precise temperatures.

Maybe, instead of just arguing as to how much the temperature reading may be out, for many, if not 99% of the people here, asking which HSF to get to keep their CPU the coolest, the built in socket thermistors is all we have to measure the temperature of our precious CPUs. If this is all we have to compare temperatures, then why can't we just state the temperatures measured by our mobos, and then other people can do the same, no matter how 'inaccurate' they are, instead of each thread turning into an arguement as to how much the temperatures are out, or compressed, or the inaccuracies as to how they are measured.

If we can just use what we have to compare our temperatures until the motherboard and CPU manufacturers come up with something better and more accurate, then can we all please just put up with what we have.



This is not, by any means, saying that the opinions of people such as Mike are not wanted here, that is NOT what I (and I'm suree, everybody here) wants. These people do lots of work to help us, but, as I've said earlier, I for one (I am yet to get a HSF which I find as good as i want) do not want every thread with people asking which HSF to get to keep their CPUs cool turning into an arguement.

 

[Mikewarrior2]

My problem with Natedog is that he thinks thinks i've made up physics.


The dangers of comparing inaccurate mb is as follows:

User A : MSI K7T Pro2a, has a 1ghz 1.8V overlcock "reading" 45C.

User B : Asus A7V 1005a Bios, Has a 1ghz 1.8V overclock "reading" Of 58C.

User C : Abit Kt7 UL Bios, has a 1ghz 1.8V overclock "reading" of 55C

You get user B wondering why his CPU temps are too high, when in fact, it isn't. Asus has a compensation built-in to the bios. But to the end-user, he doesn't know that. Assuming that the heatsink/case temp is the same(and even with the same heatsink/case temp, the above temps are possible), it adds quite a bit of confusion as to why his CPU on the A7V is running so hot.



Mike

 

[Mikewarrior2]

Or, here's another example of why comparison's are a bit mute.


User D: Msi k7t pro2, c-orb, 1.05ghz 1.9V t-bird, reading 42C.

User E: Abit Kt7, hedgehog, 1.05ghz, 1.9V t-bird, reading 45C.

User E left wondering why he spent the money for hte hedgehog when he could have "gotten" better performance with the c-orb.



Mike

 

[Liquidity]

ConfusedBW: Good topic. I've been working on my Athlon 900 o/c'd to 1Ghz and SuperOrb, attempting to determine how hot it's really getting and whether it's going to live until I'm ready to replace it. The good thing is, athlons are so cheap that if I burn this one I won't cry. It'll cost me a coupla pizzas but that's about it .

I have an ABIT KT7 motherboard with a socket thermistor. It has been reading about 100 deg F under load. I knew that was probably low so I went in and bent up the mb thermistor and put some thermal compound on the tip. Now when I rest the proc in its socket, it actually has to be pressed down (against the now bent-up thermistor) in order to be locked in place. Then I have a minimal (but enough to cover) amount of generic thermal paste on the chip where it will contact the SuperOrb. Now, under load, my CPU runs at 125 deg F (that's a stable temp after 1 hour.) Now, I've heard Mike's argument that the Orbs specifically cause socket thermistor readings to be off because they force air UNDER the socket. That sounds a little iffy to me because there is so little space for air to get under the socket. However, I'll go with it for now, and when I get home tonight I'm going to silicone the insides of the socket where it touches the motherboard. This should eliminate any erroneous readings *specific to the SuperOrb* because of airflow under the socket. I will post my findings on this thread later tonight.

 

[Mikewarrior2]

I'm not 100% sure that that is the reason for why c-orbs/s-orbs read better than they actually are.

its one theory I share with other people, people who are knowledgeable in thermodynamics. we should be testing this theory out on a p3/internal diode versus socket-thermistor shortly.

One reason that it is a plausable theory, is that something is happenening wiht c-orb/s-orbs that makes socket-thermistor readings appear far better than theoretical core temps.


Mike

 

[Liquidity]

125 deg F isn't that far off from theoretical core temps. Room temp was 68 deg F, case was open on 2 sides (top and side.)

 

[Mikewarrior2]

Yeah, you're right. One thing to watch for is that you're running 30C over ambient(reading temp wise).



Mike

 

[queyan]

MW2

You replied/answered to a question I had posted on another thread about my Vantec socket A HSF. I'd like to know if my A7V (1005a bios) is posting accurate temps using this HSF. Asus probe shows ambient at 28c and cpu temp at 53c (after playing c-strike for 2 hours). The case has 2 fans (1 in, 1 out) and is closed. I have the voltage set at 1.75v.

Thanks
Rod

 

[Mikewarrior2]

Your A7V is probably acting like hte typical A7V/Updated bios in over-approximating the temp by little bit. Chances are pretty good that your actual core temp is a bit cooler than that, so I would say you have no foreseeable heat problems either now or in the future.


Mike

 

[Matt155]

Dear fellow overclockers, MK2 is definitely right about the temperature readings of the socket thermistor. It's definitely off by
8 to 10 degrees celcius. I personally have the Abit KT7 Raid motherboard and when I was reading the temperature of my Duron 600 overclocked to 1020 mhz, at idle it was 40 C and under full load it was 45 c. So I did what the above poster did with his thermister, I bent it up so that when I inserted the cpu into the Zif socket, I had to actually press down on the cpu to clamp it flat. Now I know that the thermister is in direct contact with the bottom part of the cpu, which some have refered it as the substrate. I also know the that the thermister is actually pressing into the substrate, now with the same setup as before, my temperatures reading are 50 C idle, and 55 C under full load. I mean its common sense to me that if you have the thermister touching the cpu, one would get a more accuarate readings.
But you know what, most of these dicussions are kind of pointless because if one uses a resonable heatsink and a fan, the cpu will be operating well under Amd specs. I mean the max temp for the athons and the durons are around 190 to 200 degrees F. So I say, the temperature
readings are really only relevent to the overclocker him or herself, so we can judge what we do to the cpu is either raising or lowering the temperature of our cpu. So comparing each others temperature readings are kind of useless, since many people have various setups.

 

[Mikewarrior2]

Matt155,

When we refer to substrate, we refer to the ceramic "plate" that the cpu rests on. THe actual cpu core is the heatsink contact die, on the heatsink side of the substrate(Older socket chips did not do this, and the core was actually on the opposite side of the heatsink, which mean that socket-thermistor problems weren't as big). Even with the thermistor touching the substrate, you still only measure a percentage of cpu core temp change.

Now, with a compensated bios, like hte one that the Kt7 UL and later uses, compensated readings + a touching thermistor can usually be fairly close to "approximating" core temperature. Not always, but can be close. My recommendation has always been to use a reasonable heatsink and get a reasonable reading and everything should be great, sine AMD chips do, as you say, are capapble of high-temp operation.

Unfortunately, one issue with socket-thermistors is thermistor location. The kt7 touches substrate. the MSI/Asus boards do not. I think QDi is another that does touch substrate. The problem wtih comparing different thermistor locations is that they vary in thermal resistance from Ccore to Cback. (kt7 touch location exhibits anywhere from 37.5% of Core temp, to lwoer). Let's just say that hte MSI board measures farther out from the center of the cpu core, and the thermal resistance to that point is even higher(resulting in lower readings).


Mike

 

[Balael]

Thanks to mikes help and info, I went ahead and got a thermistor to place on the side of the core, and also just barely between the hsf and the core to get an accurate reading from my kt7-raid using the wt bios. Just to note, the only difference between touching the side of the core, and being placed between the two was only 1 degree, so i left most of the measuring to be touching the side of the core. I found that the motherboard was reading within 1 degree through most of my tests, the same temp my thermistor was reading. But one thing for sure, the older versions of the kt7 bios were in fact off on the temp readings, but the wt version seems to be accurate. The only time the two read more than a degree differently from each other, is when i put it to 1.85v, at 1080, the thermistor was reading 2 degrees less than the motherboard, so maybe too much compensation? Hope that helps some.

600 @ 600
Idle
mbm: 30
thm: 29.7
load
mbm: 31
thm: 31.1
@ 1000 w 1.75v

Idle
mbm: 36
thm: 36.9
load
mbm: 43
thm: 41.6
@1080 w 1.85v

idle
mbm: 41
thm: 40.7
load
mbm: 48
thm: 46.2
My Setup

 

[Mikewarrior2]

I wouldn't say too much compensation.

I would say just about right, since external thermsitors touching core edge read a bit lower than full core temp.

I would say the compensation loses effect at 1.1, 1.2ghz t-bird overclocks, where the +10 doesn't cover enough of the "under 50%" thermistor reading drop from core temp.


Mike